Clinical and kinematic responses to neck injuries in low-speed reverse motor vehicle collision tests: a human volunteer study

Abstract

Background

Whiplash-associated disorders (WAD) are commonly attributed to rapid cervical acceleration–deceleration during motor vehicle collisions; however, the clinical significance of low-speed reverse collisions remains uncertain.

Methods

This prospective experimental study evaluated clinical, radiological, neurophysiological, and biomechanical responses in twelve healthy adult volunteers (6 males, 6 females; mean age 31.2 ± 8.5 years) exposed to controlled reverse collisions at an impact speed of 8 km/h. Assessments included pre- and post-collision X-ray and MRI imaging, electromyography and nerve conduction studies, self-reported pain evaluation, and kinematic analysis of cervical motion, neck torque, and the Neck Injury Criterion (NIC).

Results

No structural cervical spine injuries were identified on imaging, although subtle straightening of cervical lordosis was observed in two participants. Electromyography and nerve conduction studies demonstrated statistically significant changes in selected parameters; however, all values remained within normal physiological ranges, with no findings indicative of clinically meaningful nerve injury. Self-reported pain was mild and transient, most frequently involving the cervical and lumbar regions immediately after collision and declining substantially within one week. Kinematic analysis showed that 50% of participants exceeded 30 N·m in peak neck torque, whereas all NIC values remained well below established injury thresholds.

Conclusion

In healthy adults, low-speed reverse collisions produced measurable biomechanical loads and transient musculoskeletal responses without evidence of clinically significant cervical spine injury. These findings support the use of integrated multimodal assessment—including imaging, neurophysiological testing, pain reporting, and kinematic indices—to differentiate benign biomechanical adaptations from true injury in low-velocity collision contexts. Further longitudinal studies are warranted to evaluate individual variability and the persistence of subtle cervical alignment changes.

Introduction

Whiplash-associated disorders (WAD) refer to a range of clinical symptoms resulting from the rapid acceleration–deceleration movement of the cervical spine, most commonly occurring during motor vehicle collisions (MVCs) [1]. This mechanism can result in soft tissue and neuromuscular injury, leading to pain, stiffness, and neurological symptoms [2]. Real-world crash data suggest that whiplash-associated disorders may occur even at low delta-V (ΔV), defined as the change in vehicle velocity during a collision, and that their biomechanical and clinical consequences are complex and not directly proportional to impact speed [3].

The pathophysiology of WAD involves both peripheral and central mechanisms. Tissue strain and microdamage can trigger local inflammatory responses, while sustained nociceptive input may induce central sensitization, resulting in chronic or disproportionate pain [4]. Because radiological findings such as MRI or CT scans often fail to identify clear structural abnormalities, diagnosis frequently depends on subjective symptom reporting. This diagnostic challenge has contributed to controversy regarding the legitimacy of reported symptoms following low-speed collisions [5].

The debate over whether low-velocity or light-contact accidents can cause genuine neck injuries remains active in both scientific and social contexts. In many countries, including South Korea, whiplash-related claims have drawn attention due to the difficulty in objectively verifying injury severity and the consequent increase in medical and insurance costs. For instance, treatment costs for whiplash and neck sprains caused by rear-end collisions have been estimated at approximately USD 2.7 billion annually in the United States and 10 billion euros in Europe [6, 7]. In South Korea, as of 2022, 3,629 vehicle-to-vehicle collisions occurred while driving in reverse, resulting in 4,907 injuries—an increase of over 40% compared to 2018 [8]. Many of these cases involve minor soft-tissue injuries that are difficult to assess objectively, complicating compensation and insurance evaluations.

Previous biomechanical studies have sought to establish the threshold at which neck injury becomes biomechanically plausible. Castro et al. (2014) reported that hyperextension of the neck did not occur in volunteer experiments with ΔV values between 8.3 and 14.2 km/h, proposing a “limit of harmlessness” around 10–15 km/h [9]. Brault et al. (1998) demonstrated that headrest positioning significantly affects cervical motion during rear-end impacts, while Szabo et al. (2009) found that pre-existing conditions could increase vulnerability even at low ΔV levels [10, 11]. In Korea, a controlled study involving 50 male volunteers aged 30–50 years in simulated rear-end collisions (ΔV: 5–8 km/h) reported no abnormal findings or complaints of discomfort [12]. Nevertheless, the biomechanical threshold for neck strain in minor MVCs remains inconclusive, especially for collisions in reverse.

Despite these efforts, limited research has investigated the association between ΔV and actual clinical responses in low-speed reverse collisions, particularly within Asian populations. Understanding this relationship is critical for distinguishing biomechanically substantiated injuries from psychosocially influenced symptom reporting.

Therefore, the present study aims to investigate subclinical clinical, neurophysiological, and kinematic responses in healthy human volunteers exposed to low-speed reverse motor vehicle collisions that are generally considered non-injurious. By integrating biomechanical and clinical assessments, this study seeks to provide objective reference data that may help distinguish benign biomechanical adaptations from true cervical injury under low-velocity collision conditions.

Methods

Study design and participants

This study was designed as a prospective experimental trial to evaluate human kinematic and clinical responses during low-speed rear-end collisions while driving in reverse. Twelve healthy adults (6 males and 6 females; mean age = 31.2 ± 8.5 years; range = 21–47 years) with valid driver’s licenses were recruited. Participants were selected to represent both sexes and a range of adult ages to explore potential demographic differences in biomechanical and clinical responses. Exclusion criteria included prior cervical or lumbar spine disorders, neurological diseases, or any history of insurance claims related to neck or back injuries, to minimize confounding factors and health risks.

All participants provided written informed consent. The study was approved by the Institutional Review Board of Yonsei University Wonju Severance Christian Hospital (IRB No. CR323046) and was conducted in accordance with the International Council for Harmonisation (ICH) Good Clinical Practice (E6[R2]) guidelines and the principles of the Declaration of Helsinki. It also complied with the Korean Good Clinical Practice (KGCP) and relevant regulations, with due consideration for the rights and safety of the participants. As this investigation did not fall under the definition of a clinical trial, registration was not applicable.

Experimental setup and collision scenarios

To simulate realistic low-speed reverse collisions, six vehicle models representing different mass categories were used: two compact sedans (Spark LT^®, Kona OS^®), two large sedans (Genesis DH^®, Grandeur IG^®), and two SUVs (Sorento UM^®, BMW X4^®). Each vehicle combination (e.g., Sorento UM^® versus Spark LT^®) was tested six times at a fixed impact speed of 8 km/h, with the front vehicle backing into the rear vehicle while the rear vehicle was stationary. This value was chosen based on a previous volunteer safety trial that reported no structural or neurological damage at speeds below 10 km/h [9]. This speed allowed controlled reproduction of minor collision dynamics while ensuring participant safety. To examine possible geometric and inertial effects, collision configurations alternated between 100% frontal contact and 50% offset. Driver and rear-seat passenger positions were switched between trials to minimize order bias and assess positional variability (Table 1).

Table 1.

Low-speed collision scenarios when driving in reverse

Test No,	Vehicle model		Speed (㎞/h)	Offset (%)	Delta V (㎞/h)	Participant No. (Sex/Age)
	Colliding	Collided
						FD seat	RR seat
T1	Grandeur IG	Genesis DH	8.0	100	4.6	P1 (F/30)	P2 (M/32)
T2	Genesis DH	Grandeur IG	8.0	50	5.2	P3 (M/29)	P4 (F/22)
T3	Sorento UM	Spark LT	8.0	100	7.8	P5 (F/36)	P6 (M/47)
T4	Spark LT	Sorento UM	8.0	50	3.7	P7 (M/47)	P8 (F/30)
T5	X4	Kona OS	8.0	100	6.1	P9 (F/29)	P10 (M/28)
T6	Kona OS	X4	8.0	50	3.8	P11 (M/21)	P12 (F/24)

* T: test, P: participant, FD: front driver, RR: rear right, M: male, F: female

Participants maintained a standardized driving posture—hands placed on the steering wheel and back against the seat—to mimic real-world driving conditions. To minimize anticipatory muscle tension, participants wore eye masks and headphones during collisions. A camera positioned laterally captured upper body motion, and participants with long hair tied it up to ensure visibility of cervical landmarks (Fig. 1).

Fig. 1.

Test conditions: a driver seat, b 2nd-row right rear seat, c camera installation

Radiological evaluation

Comprehensive pre- and post-collision imaging was performed using X-ray and MRI to assess structural and soft tissue changes. X-ray (C-spine AP, lateral, oblique, flexion, and extension views) evaluated bony alignment and curvature. The cervical lordotic angle was quantified using the Cobb method between C2 and C7. MRI (performed within 48 h post-impact to ensure consistency) included T2 Dixon sagittal, T2 axial, GRE axial, and DTI sequences, focusing on ligamentous, muscular, and neural integrity. Quantitative MRI parameters such as signal intensity ratio (SIR) of paraspinal muscles and prevertebral fluid thickness (mm) were extracted to supplement visual assessments. Soft tissue analysis extended to paraspinal muscles, ligaments, and the identification of prevertebral fluid collections. Within the images, the radiologist paid particular attention to ligaments, assessing their integrity and searching for any signs of injury that might impact spinal stability. Evaluation of alignment and position changes, including instances of straightening, offered additional insights into potential muscle spasms or underlying pathologies. Additionally, to calculate the moment of inertia at the neck joint, sagittal cervical MRI images were analyzed to derive subject-specific neck length, defined as the distance from the estimated center of rotation at the lower cervical spine to the head center of mass. The neck–head segment was modeled as a rigid body, and the moment of inertia was estimated using anthropometric assumptions based on segment length and mass distribution. In this study, participants’ neck length was measured, defined as the distance between the upper part of the hyoid bone to the jugular notch [13]. Radiologists were blinded to participant identity, test sequence, and symptom reports to prevent interpretation bias.

Electromyography (EMG) and Nerve Conduction Velocity (NCV) test

To assess neuromuscular function and peripheral nerve integrity in the cervical region, Electromyography (EMG) and Nerve Conduction Velocity (NCV) tests were conducted both prior to and precisely 72 h after the collision test. This standardized interval was chosen to minimize temporal variability in electrophysiological recovery. As part of a comprehensive evaluation alongside X-ray and MRI imaging, these neurophysiological tests aimed to detect collision-induced changes in cervical spine function. EMG recordings captured resting potentials, motor unit action potential (MUAP) amplitudes, and recruitment patterns, including spontaneous activity and altered firing sequences indicative of muscle injury or tension [14]. NCV analysis measured conduction velocity, CMAP (Compound Muscle Action Potential), and SNAP (Sensory Nerve Action Potential) across the median, ulnar, and cervical spinal nerves to evaluate the speed and integrity of electrical impulse transmission. Abnormal findings such as delayed latency or reduced CMAP/SNAP amplitudes were flagged as potential signs of cervical radiculopathy or peripheral polyneuropathy. All electrodiagnostic data were interpreted by rehabilitation specialists who were blinded to imaging and pain assessment results, ensuring unbiased analysis. Their review focused on identifying neurophysiological abnormalities that could signal post-impact nerve dysfunction or muscular compromise.

Pain and symptom assessment

Self-reported pain was assessed at three standardized time points: immediately after the collision, 24 h later, and 7 days post-impact. Participants were asked to indicate whether pain occurred and to rate its intensity using the Numeric Rating Scale (NRS; 0–10), a validated tool for quantifying subjective pain levels [15]. Up to three affected body regions were selected from the cervical spine (neck), shoulder and upper arm, back, lumbar spine (lower back), and waist. The frequency, intensity, and anatomical distribution of reported pain were analyzed across time points to identify temporal patterns and potential delayed onset symptoms.

Extraction of high-resolution motion data and kinematic analysis

To analyze the dynamic response of the head–neck–trunk system during collision, high-resolution motion data were extracted using a markerless 3D motion capture system (REMOBODY-S^®, REMO Inc., Republic of Korea). The system implements a deep-learning–based 3D human pose estimation network derived from the metric-scale, truncation-robust Metrabs framework, which regresses absolute 3D joint coordinates from monocular RGB images in the camera coordinate system [16]. From each video frame, 2D keypoints are first detected, and the corresponding 3D pose is then estimated in metric space; the resulting joint coordinates are subsequently used to compute posture-related parameters such as cervical angles and joint torques.

To evaluate the measurement accuracy of the markerless system against a marker-based gold standard, we validated the model using the CMU Panoptic Motion Capture dataset [17], which provides synchronized multi-view RGB images and marker-based 3D motion capture data (Vicon system) as ground truth. The consistency between the markerless estimates and the Vicon-based reference data was quantified using the mean per-joint position error (MPJPE), defined as the average Euclidean distance between corresponding joint positions [18]. In this evaluation, the markerless system achieved an MPJPE of 32.73 mm across 14 joints relative to the Vicon ground truth. This performance was independently confirmed in an external AI voucher performance test conducted by an accredited testing laboratory (STA Testing & Consulting, Seoul, Republic of Korea), which reported the same MPJPE value of 32.73 mm for the deployed model configuration. These error levels are comparable to those reported for state-of-the-art markerless 3D human pose estimation methods [16, 18] and are sufficient for quantifying cervical kinematics and joint torques under the low-speed collision conditions examined in this study.

The overall data-processing workflow of the markerless system can be summarized as follows: (1) extraction of RGB image sequences from lateral collision videos; (2) detection of the target person and 2D skeletal keypoints in each frame; (3) regression of absolute 3D joint coordinates in the camera coordinate system using the pretrained deep-learning model; (4) temporal smoothing of joint trajectories; and (5) computation of segment orientations, joint angles, angular velocities and accelerations, and net joint torques based on the reconstructed 3D joint coordinates. Subsequent steps for constructing segment-based coordinate systems, estimating cervical joint torques via Newton–Euler dynamics, and calculating the Neck Injury Criterion (NIC) are described in detail below.

This system reconstructed human keypoints in the Skinned Multi-Person Linear Model (SMPL+) format, including the head segment (SMPL+head) from 2D camera images [19, 20]. Segment orientations were computed by establishing anatomically defined local coordinate systems based on key landmarks. For each body segment—trunk, neck, and head—orthonormal coordinate axes were constructed using vector relationships and cross products to ensure right-handedness and continuity. The trunk orientation was derived from shoulder and neck base points, while the neck and head orientations incorporated vectors between the neck base, upper neck, and ears. These matrices enabled calculation of relative motion angles between segments using 3D Euler rotations, capturing flexion-extension, lateral bending, and axial rotation. Linear and angular velocities were derived from filtered positional data using an 8th -order low-pass Butterworth filter (cutoff = 10 Hz). Angular accelerations and joint torques were estimated via Newton–Euler dynamics to model inertial loads on the upper (C1) and lower (C7/T1) cervical joints. To calculate the moment of inertia used in the joint torque estimation, sagittal cervical MRI images were analyzed to derive subject-specific neck length, defined as the distance between the estimated centers of rotation at the upper cervical (C1) and lower cervical (C7/T1) levels and the head center of mass. The head–neck segment was modeled as a rigid body, and its moment of inertia was estimated using standard anthropometric assumptions based on segment length and mass distribution. Angular accelerations were obtained from kinematic motion data, and joint torques at C1 and C7/T1 were calculated using inverse Newton–Euler dynamics, incorporating rotational inertia and measured angular acceleration (τ = I·α). Additionally, the Neck Injury Criterion (NIC) was computed based on relative head–thorax acceleration and velocity to assess potential injury risk.

Statistical analysis

All statistical analyses were performed using IBM SPSS Statistics Version 27.0 (IBM Corp., Armonk, NY, USA). Electrophysiological parameters obtained from EMG and NCV tests were analyzed using two-way repeated measures ANOVA with side (left vs. right) and time (pre- vs. post-collision) as within-subject factors. When significant main effects or interaction effects were observed, post-hoc pairwise comparisons were conducted with Bonferroni adjustment to control for type I error. Assumptions of sphericity were tested using Mauchly’s test, and when violated, the Greenhouse-Geisser correction was applied. Partial eta-squared (η²) values were calculated to estimate effect sizes. A p-value of less than 0.05 was considered statistically significant. Kinematic variables, including peak angular velocity, neck torque, and the Neck Injury Criterion (NIC), were compared between pre- and post-collision conditions using paired t-tests. Radiological and MRI-derived parameters from cervical spine MRI were evaluated by board-certified radiologists who independently classified each scan as either positive or negative for structural abnormalities. Self-reported pain data were analyzed across three time points: immediately post-collision, 1 day after, and 1 week after. Pain intensity was quantified using the Numeric Rating Scale (NRS; 0–10), and descriptive statistics (mean ± standard deviation) were calculated for each time point. The number of participants reporting pain in each anatomical region (cervical spine, shoulder/upper arm, lumbar spine, waist) was summarized using frequency counts and proportions. Because participants could report pain in multiple regions simultaneously, region-specific pain frequencies were treated as independent binary variables.

Results

X-ray images and MRI readings before and after collision tests

Table 2 shows that cervical spine MRI conducted before and after the low-speed rear-end collision revealed no evidence of structural injury across all participants. Specifically, there were no detectable fractures, bone marrow edema, spinal cord signal intensity abnormalities, or soft tissue damage including paraspinal muscle changes, prevertebral fluid collection, or ligament disruption. However, a notable finding emerged with two out of twelve participants displaying a subtle straightening of the cervical spine post-collision, suggesting a potential biomechanical impact on spinal alignment. These findings suggest that even in the absence of radiologically apparent injuries, subtle musculoskeletal responses such as alignment straightening may occur post-collision, highlighting the importance of dynamic and functional assessment in mild cervical trauma.

Table 2.

Pre- and post-collision cervical spine MRI findings

Participant	Fracture	Bone Marrow Edema	Cord Signal Intensity	Paraspinal Muscle	Prevertebral Fluid	Ligament Injury	Cervical Alignment Straightening
	Pre / Post	Pre / Post	Pre / Post	Pre / Post	Pre / Post	Pre / Post	Pre / Post
P1	– / –	– / –	– / –	– / –	– / –	– / –	+ / +
P2	– / –	– / –	– / –	– / –	– / –	– / –	+ / +
P3	– / –	– / –	– / –	– / –	– / –	– / –	+ / +
P4	– / –	– / –	– / –	– / –	– / –	– / –	+ / +
P5	– / –	– / –	– / –	– / –	– / –	– / –	+ / +
P6	– / –	– / –	– / –	– / –	– / –	– / –	– / +
P7	– / –	– / –	– / –	– / –	– / –	– / –	– / –
P8	– / –	– / –	– / –	– / –	– / –	– / –	+ / +
P9	– / –	– / –	– / –	– / –	– / –	– / –	+ / +
P10	– / –	– / –	– / –	– / –	– / –	– / –	– / +
P11	– / –	– / –	– / –	– / –	– / –	– / –	+ / +
P12	– / –	– / –	– / –	– / –	– / –	– / –	+ / +

* (+): Positive finding, (-): No abnormality

EMG and NCS studies before and after collision tests

Table 3 summarizes the motor and sensory nerve conduction study results for the different nerve segments tested. Repeated measures ANOVA was conducted to assess the main effects of Time (pre- vs. post-collision), Side (left vs. right), and their interaction (Time × Side) on nerve conduction parameters. For the median motor nerve (Wrist-APB), a significant main effect of Time was observed for latency (F(1,11) = 14.211, p = 0.007, η²=0.670), and for the median motor nerve (Elbow-Wrist), conduction velocity also showed a significant Time effect (F(1,11) = 15.446, p = 0.006, η²=0.688). Additionally, a significant Side effect was found for duration (F(1,11) = 7.692, p = 0.028, η²=0.524). For the ulnar motor nerve (Wrist-ADM), significant Time effects were identified for peak amplitude (F(1,11) = 5.778, p = 0.047, η²=0.452) and area (F(1,11) = 6.257, p = 0.041, η²=0.472), with Time × Side interactions for latency (F(1,11) = 8.549, p = 0.022, η²=0.550) and onset amplitude (F(1,11) = 8.464, p = 0.023, η²=0.547). No significant changes were observed for the ulnar motor nerve (Below Elbow-Wrist) or the median sensory nerve (Finger-Wrist) (all p > 0.05). For the ulnar sensory nerve (Wrist-Finger), onset latency (F(1,11) = 7.569, p = 0.020, η²=0.431) and peak latency (F(1,11) = 10.563, p = 0.009, η²=0.514) showed significant Time effects. Although some nerve conduction parameters demonstrated statistically significant changes, all values remained within established normal reference ranges for latency, amplitude, and conduction velocity as defined by standard electrodiagnostic guidelines. No pathological patterns indicative of nerve injury were observed. Importantly, no consistent pattern of abnormality or asymmetry suggestive of nerve dysfunction or cervical spine damage was observed following the low-speed rear-end collision. These findings support the conclusion that the collision conditions studied did not result in cervical spine injury detectable by nerve conduction studies.

Table 3.

Summary of repeated measures two-way ANOVA results for nerve conduction parameters before and after the collision test (main effects of Time, Side, and Time × Side interaction)

Nerve conduction parameters	Variables
	Time			Side			Time*Side
	F (df1,df2)	p	η²	F (df1,df2)	p	η²	F (df1,df2)	p	η²
Median Motor (Wrist-APB)
Latency (ms)	14.211 (1,7)	0.007‡	0.670	0.766 (1,7)	0.410	0.099	0.120 (1,7)	0.739	0.017
Amplitude (mV, onset)	0.507 (1,7)	0.499	0.068	0.234 (1,7)	0.643	0.032	2.496 (1,7)	0.158	0.263
Amplitude (mV, peak)	1.784 (1,7)	0.223	0.203	1.042 (1,7)	0.341	0.130	2.887 (1,7)	0.133	0.292
Stimulus Intensity (mA)	0.309 (1,7)	0.596	0.042	0.008 (1,7)	0.933	0.001	0.024 (1,7)	0.882	0.003
Duration (ms)	0.028 (1,7)	0.872	0.004	0.843 (1,7)	0.389	0.107	0.069 (1,7)	0.801	0.010
Area (ms*mV)	2.926 (1,7)	0.131	0.295	0.068 (1,7)	0.802	0.010	0.594 (1,7)	0.466	0.078
Median Motor (Elbow-Wrist)
Latency (ms)	1.052 (1,7)	0.339	0.131	0.034 (1,7)	0.860	0.005	0.904 (1,7)	0.373	0.114
Amplitude (mV, onset)	1.228 (1,7)	0.304	0.149	0.794 (1,7)	0.402	0.102	0.869 (1,7)	0.382	0.110
Amplitude (mV, peak)	2.836 (1,7)	0.136	0.288	2.317 (1,7)	0.172	0.249	1.462 (1,7)	0.266	0.173
Conduction Velocity (ms)	15.446 (1,7)	0.006‡	0.688	0.751 (1,7)	0.415	0.097	1.861 (1,7)	0.215	0.210
Distance (mm)	1.000 (1,7)	0.351	0.125	0.467 (1,7)	0.516	0.063	1.000 (1,7)	0.351	0.125
Stimulus Intensity (mA)	2.460 (1,7)	0.161	0.260	0.778 (1,7)	0.407	0.100	0.186 (1,7)	0.679	0.026
Duration (ms)	0.336 (1,7)	0.580	0.046	7.692 (1,7)	0.028‡	0.524	1.358 (1,7)	0.282	0.163
Area (ms*mV)	3.321 (1,7)	0.111	0.322	0.021 (1,7)	0.889	0.003	0.568 (1,7)	0.476	0.075
Ulnar Motor (Wrist-ADM)
Latency (ms)	2.238 (1,7)	0.178	0.242	5.437 (1,7)	0.052	0.437	8.549 (1,7)	0.022‡	0.550
Amplitude (mV, onset)	0.452 (1,7)	0.523	0.061	0.191 (1,7)	0.675	0.027	8.464 (1,7)	0.023‡	0.547
Amplitude (mV, peak)	5.778 (1,7)	0.047‡	0.452	0.848 (1,7)	0.388	0.108	1.381 (1,7)	0.278	0.165
Stimulus Intensity (mA)	0.006 (1,7)	0.943	0.001	0.028 (1,7)	0.871	0.004	0.221 (1,7)	0.653	0.031
Duration (ms)	0.006 (1,7)	0.942	0.001	2.166 (1,7)	0.185	0.236	0.046 (1,7)	0.836	0.007
Area (ms*mV)	6.257 (1,7)	0.041‡	0.472	0.135 (1,7)	0.725	0.019	0.002 (1,7)	0.963	0.000
Ulnar Motor (Bl.Elbow-Wrist)
Latency (ms)	0.082 (1,7)	0.783	0.012	0.270 (1,7)	0.619	0.037	0.425 (1,7)	0.535	0.057
Amplitude (mV, onset)	1.052 (1,7)	0.339	0.131	0.047 (1,7)	0.835	0.007	0.114 (1,7)	0.745	0.016
Amplitude (mV, peak)	4.533 (1,7)	0.071	0.393	0.027 (1,7)	0.874	0.004	0.065 (1,7)	0.805	0.009
Conduction Velocity (ms)	2.175 (1,7)	0.184	0.237	0.502 (1,7)	0.501	0.067	1.956 (1,7)	0.205	0.218
Stimulus Intensity (mA)	0.632 (1,7)	0.453	0.083	0.121 (1,7)	0.739	0.017	0.136 (1,7)	0.724	0.019
Duration (ms)	0.382 (1,7)	0.556	0.052	0.384 (1,7)	0.555	0.052	0.327 (1,7)	0.585	0.045
Area (ms*mV)	5.317 (1,7)	0.055	0.432	0.326 (1,7)	0.586	0.045	0.037 (1,7)	0.853	0.005
Median Sensory (Finger-Wrist)
Latency (Onset) (ms)	0.556 (1,11)	0.471	0.048	0.743 (1,11)	0.407	0.063	0.801 (1,11)	0.390	0.068
Latency (Peak) (ms)	0.516 (1,11)	0.487	0.045	0.856 (1,11)	0.375	0.072	0.719 (1,11)	0.415	0.061
Amplitude (mV)	0.001 (1,11)	0.973	0.000	0.408 (1,11)	0.536	0.036	0.131 (1,11)	0.724	0.012
Stimulus Intensity (mA)	0.322 (1,11)	0.582	0.028	0.235 (1,11)	0.637	0.021	0.738 (1,11)	0.408	0.063
Ulnar Sensory (Wrist-Finger)
Latency (Onset) (ms)	7.569 (1,10)	0.020‡	0.431	0.053 (1,10)	0.823	0.005	0.700 (1,10)	0.422	0.065
Latency (Peak) (ms)	10.563 (1,10)	0.009‡	0.514	0.491 (1,10)	0.499	0.047	0.140 (1,10)	0.717	0.014
Amplitude (mV)	0.036 (1,10)	0.854	0.004	1.718 (1,10)	0.219	0.147	1.769 (1,10)	0.213	0.150
Stimulus Intensity (mA)	0.142 (1,10)	0.714	0.014	4.206 (1,10)	0.067	0.296	0.891 (1,10)	0.368	0.082

^‡Greenhouse-Geisser corrected values reported (p<0.05)

* APB: abductor pollicis brevis, ADM: abductor digiti minimi, Bl.: below, ms: millisecond, mV: millivolt, mA: milliampere, η²: partial eta squared

Self-reported pain assessment

Table 4 presents the results of a questionnaire assessing participants’ pain perception following the collision test at three different time points: immediately after the collision, one day after, and one week after. Immediately after the collision, one participant reported pain in the cervical spine with an average pain level of 2.11 ± 1.17 on the Numeric Rating Scale (NRS), while one participant also reported pain in the shoulder and upper arm. Additionally, five participants experienced pain in the lumbar spine and two in the waist area. At this time, five participants reported no pain. One day after the collision, no participants reported cervical spine pain, although the average pain level remained at 2.25 ± 1.89; one participant reported pain in the shoulder and upper arm, and three reported lumbar spine pain, with no reports of waist pain. Eight participants reported no pain at this time. One week after the collision, no participants reported cervical spine pain. One participant reported shoulder or upper limb pain, and one reported lumbar pain, while the number of participants reporting no pain increased to ten. It is important to note that participants could report pain in multiple body regions simultaneously. Overall, pain reports decreased over time, particularly in the cervical spine, while the number of participants without pain increased, and the pain intensity remained relatively low throughout the observation period.

Table 4.

Results of a questionnaire on participants’ pain perception after the collision test

Time of survey	Body region with pain	No. of participants reporting pain	Level of pain using Numeric Rating Scale	No. of participants with no pain
Immediately after collision test	Cervical spine	1	2.11 ± 1.17 point	5
	Shoulder and upper arm	1
	Lumbar spine	5
	Waist	2
One day after collision test	Cervical spine	0	2.25 ± 1.89 point	8
	Shoulder and upper arm	1
	Lumbar spine	3
	Waist	0
One week after collision test	Cervical spine	0	N/A	10
	Shoulder and upper arm	1
	Lumbar spine	1
	Waist	0

* Available responding to multiple pain body regions

Biomechanical assessment of cervical spine kinematics and metrics during occupant motion

Table 5 presents the anthropometric characteristics and cervical response metrics of twelve participants exposed to low-speed rear-end collisions. Participants exhibited a range of body dimensions, with height varying from 1.58 m to 1.81 m and weight ranging from 42 kg to 90 kg. Neck length ranged from 0.106 m to 0.138 m, while neck circumference ranged from 0.316 m to 0.392 m. Peak neck torque measurements spanned from 16.58 N·m (P8) to 50.80 N·m (P10). Among the twelve participants, six (P3, P4, P5, P9, P10, and P11) recorded torque values exceeding 30 N·m. The remaining six participants (P1, P2, P6, P7, P8, and P12) showed torque values below this threshold. The NIC values ranged between 0.58 m²/s² (P8) and 2.47 m²/s² (P6). All participants exhibited NIC values below 15 m²/s². No participant exceeded this threshold, indicating consistency in NIC responses across the group. Overall, half of the participants (6 out of 12) exceeded the 30 N·m torque threshold, whereas none exceeded the 15 m²/s² NIC threshold. This dichotomy suggests a greater variation in torque response relative to NIC values across individuals within the same exposure condition.

Table 5.

Participant anthropometry and cervical injury metrics: peak torque and NIC with threshold-based evaluation

Participant	Height (m)	Weight (kg)	Neck Length (m)	Neck Circumstance (m)	Peak Neck Torque (N*m)	NIC (m2/s2)	Torque Threshold (> 30 N*m)	NIC Threshold (> 15 m2/s2)
P1	1.61	42	0.129	0.317	24.91	1.32	No	No
P2	1.81	78	0.124	0.378	22.96	1.38	No	No
P3	1.77	75	0.136	0.383	33.45	1.53	Yes	No
P4	1.67	57	0.138	0.321	39.51	2.22	Yes	No
P5	1.70	63	0.130	0.321	33.33	2.27	Yes	No
P6	1.73	72	0.127	0.392	28.05	2.47	No	No
P7	1.78	75	0.118	0.392	29.38	0.89	No	No
P8	1.66	53	0.132	0.321	16.58	0.58	No	No
P9	1.58	90	0.110	0.324	39.08	1.64	Yes	No
P10	1.69	70	0.124	0.381	50.80	1.78	Yes	No
P11	1.69	61	0.106	0.376	43.38	0.82	Yes	No
P12	1.61	53	0.114	0.316	27.39	1.05	No	No

* NIC: Neck injury criterion

Discussion

This study investigated clinical, radiological, neurophysiological, and biomechanical responses to low-speed (~ 8 km/h) reverse collisions in 12 healthy adult volunteers. By integrating multimodal outcome measures, including imaging, neurophysiological testing, subjective pain assessment, and quantitative kinematic indices, this study directly addresses ongoing clinical and biomechanical controversies regarding the injury potential of low-speed collisions. While no overt structural injuries were detected, several subtle physiological and mechanical responses were observed, suggesting that even low-velocity impacts may induce transient adaptations in the cervical spine and surrounding structures without constituting clinically meaningful injury. This distinction is particularly important in contexts where low-speed collisions are frequently associated with diagnostic uncertainty and heterogeneous clinical interpretation. Given the limited sample size and demographic homogeneity, these findings should be interpreted cautiously and not generalized to broader populations.

Radiological findings and cervical alignment

Pre- and post-collision MRI assessments of the cervical spine in 12 healthy participants revealed no evidence of fractures, bone marrow edema, spinal cord signal abnormalities, or soft tissue damage, including ligament disruption. However, 2 out of 12 participants (16.7%) exhibited subtle straightening of the cervical lordosis following the low-speed reverse collision. Although overt structural injury was absent, this alignment change may reflect an acute musculoskeletal response, such as reflexive muscle guarding or postural compensation. Biomechanical models of cervical acceleration–deceleration (CAD) injuries describe an initial “snap-through” or S-shaped buckling of the spine during impact, which can transiently alter sagittal curvature [21]. Previous longitudinal MRI research has also demonstrated that degenerative changes in the cervical spine may progress over decades, though not necessarily correlate with clinical symptoms [22]. Our acute post-impact findings suggest that cervical straightening may represent an early biomechanical adaptation rather than a direct indicator of pathology. While our study did not include post-collision interventions, the presence of alignment changes (even in low-velocity impacts) underscores the importance of evaluating sagittal curvature in post-trauma imaging protocols. Furthermore, ligamentous injury has been hypothesized as a contributor to persistent lordotic loss. Studies using high-resolution MRI and radiography have reported transverse ligament changes and sustained curve flattening even after muscle spasms resolve [23, 24]. This supports the notion that cervical straightening may not be purely transient but could reflect enduring soft tissue strain or micro-instability. Given its established association with poorer clinical outcomes, further research is warranted to assess the persistence of lordotic loss, its relationship to emerging symptoms, and the potential benefits of early therapeutic interventions such as cervical extension traction.

Neurophysiological observations

Neurophysiological assessments revealed statistically significant changes in selected nerve conduction parameters following the low-speed reverse collision. For instance, median motor nerve latency (wrist-APB) increased significantly (F(1,11) = 14.211, p = 0.007, η²=0.670), and median motor nerve conduction velocity (elbow-wrist) decreased (F(1,11) = 15.446, p = 0.006, η²=0.688). However, these changes remained within normal physiological limits and showed no pathological pattern suggestive of nerve injury. These findings are consistent with previous studies that reported no neurophysiological evidence of cervical spine injury in low-speed rear-end collision scenarios, despite subjective symptom reports [25]. Other studies have also indicated that cervical muscle activity changes may occur in response to minor collisions, but without corresponding nerve damage [26].

The timing of post-collision assessments represents an important contextual factor in interpreting the present findings. MRI was performed within 48 h after impact to capture potential acute soft-tissue or ligamentous changes, while EMG and nerve conduction studies were conducted at 72 h to assess early neuromuscular responses after the resolution of immediate transient effects. This assessment window aligns with commonly used clinical evaluation timeframes for detecting early post-traumatic abnormalities. Nevertheless, it is acknowledged that ultra-acute physiological responses occurring within the first few hours after collision—such as short-lived muscle activation changes or transient inflammatory responses—may not have been fully captured. Importantly, however, the absence of structural abnormalities on early MRI and the persistence of neurophysiological parameters within normal ranges at 72 h suggest that no clinically meaningful or evolving cervical pathology developed during the early post-impact period [14].

Our results reinforce the notion that statistically significant variations in nerve conduction parameters do not necessarily equate to clinically meaningful injury. Future studies employing multi-timepoint designs, including ultra-acute and longer-term follow-up assessments, would be valuable in further characterizing the temporal evolution of neuromuscular responses following low-speed collisions. Importantly, interpretations were made without attributing symptoms to non-physiological motives such as compensation-seeking, in accordance with principles of scientific neutrality and objectivity.

Pain perception and unexpected lumbar symptoms

Immediately following the low-speed reverse collision, 58.3% of participants reported experiencing pain, with primary complaints localized to the cervical spine (71.4% of those with pain) and lower back (42.9%). One week later, the overall pain prevalence declined to 25.0%, with symptoms persisting mainly in the cervical region (66.7% of those with persistent pain). The average pain intensity remained low throughout the observation period, and no participants reported persistent or escalating symptoms. These findings are broadly consistent with previous studies, including a Korean low-speed collision cohort in which 12% of subjects reported transient neck discomfort at ΔV 5–8 km/h [27]. The slightly higher immediate pain report rate in our study may reflect differences in collision direction (reverse vs. rear-end), participant awareness, or experimental context.

Notably, nearly half of the participants reported lumbar pain immediately post-collision, despite the cervical spine being the primary site of mechanical loading. This unexpected distribution may be explained by several plausible mechanisms: trunk–pelvis kinematic coupling during impact, seat design and posture effects, particularly trunk rotation during reverse driving, reflexive muscle co-contraction, or heightened attentional focus due to experimental awareness. Among these, posture-related factors and biomechanical coupling appear most plausible in our cohort, given the absence of structural lumbar findings.

Two limitations should be acknowledged. First, although lumbar pain was reported more frequently than cervical pain, objective biomechanical or imaging data of the lumbar spine were not collected, limiting mechanistic interpretation. Second, subjective pain reporting may have been influenced by psychological expectations or experimental awareness. Future studies incorporating lumbar biomechanical assessment and psychological profiling are warranted.

Kinematic and biomechanical analysis

In our test cohort, 50% of participants exceeded 30 N·m in peak neck torque during low-speed reverse collisions, yet no clinical or radiological injuries were observed. This finding parallels cadaveric studies in which transient torque spikes failed to produce tissue damage, suggesting that torque alone may not reliably predict injury risk in brief, low-velocity exposures [28]. Torque represents rotational force, which can vary significantly depending on individual anatomy, neck length, and posture. In contrast, the Neck Injury Criterion (NIC), which integrates translational acceleration and velocity between the head and thorax, offers a broader measure of inertial loading. All NIC values in our study ranged from 0.58 to 2.47 m²/s², remaining well below the commonly referenced injury threshold of 15 m²/s² [29, 30]. Some biomechanical models have proposed even lower thresholds (~ 8.7 m²/s²) for detecting intervertebral damage [31], yet our data consistently fell within a biomechanically tolerable range.

An important consideration in interpreting these findings is the variability in ΔV observed across collision trials, despite a fixed vehicle impact speed of 8 km/h. Differences in vehicle mass, geometry, and stiffness resulted in ΔV values ranging from 3.7 to 7.8 km/h, reflecting the conditions commonly encountered in real-world low-speed collisions. Although ΔV represents the effective biomechanical load transmitted to the occupant, no systematic relationship was observed between ΔV magnitude and clinical symptoms, neurophysiological alterations, or biomechanical injury indicators in the present study. Notably, across this ΔV spectrum, all neurophysiological parameters remained within established normal ranges, no radiological injuries were detected, and NIC values consistently remained well below reported injury thresholds. These findings suggest that, within the observed ΔV range, low-speed reverse collisions are unlikely to produce clinically meaningful cervical injury in healthy adults [32, 33].

The observed discrepancy between elevated torque values and low NIC scores underscores the importance of multi-metric evaluation when assessing cervical spine injury potential. This finding challenges the clinical tendency to infer injury risk from isolated biomechanical parameters and highlights the need for integrated interpretive frameworks. Relying solely on a single biomechanical parameter may either overestimate or underestimate actual risk. By demonstrating that elevated torque can occur in the absence of pathological imaging, abnormal neurophysiology, or injurious NIC levels, the present study provides objective evidence that helps reconcile biomechanical measurements with clinical outcomes. Future research should explore whether torque thresholds correlate with subjective symptoms, musculoskeletal adaptations such as cervical straightening, or long-term outcomes. Expanding biomechanical models to account for individual variability and combining rotational and translational metrics may enhance predictive accuracy in low-speed collision studies.

Conclusion

This study investigated clinical, radiological, neurophysiological, and biomechanical responses to low-speed (~ 8 km/h) reverse collisions in healthy adults, with the hypothesis that such impacts do not result in cervical spine injury. Across multimodal assessments, including high-resolution MRI, EMG/NCS testing, subjective pain scales, and kinematic indices such as NIC and torque, no clinically significant injuries were detected. Observed changes, such as cervical spine straightening and transient pain, appear to reflect acute-phase musculoskeletal adaptations rather than structural damage or long-term pathology.

Importantly, all neurophysiological variations remained within normal limits, and NIC values fell well below established injury thresholds, reinforcing the biomechanical tolerability of low-speed reverse impacts in healthy individuals. These findings provide empirical evidence supporting the use of objective, multimodal evaluation protocols to distinguish benign biomechanical responses from true injury. This distinction is particularly relevant in clinical and medicolegal contexts, where low-speed collisions often generate disproportionate diagnostic ambiguity and variability in management decisions.

However, the long-term implications of these subtle changes remain uncertain. Given the limited sample size and demographic homogeneity, results should not be generalized to older adults or individuals with pre-existing musculoskeletal conditions. Future research should incorporate longitudinal follow-up, therapeutic intervention trials, and expanded cohorts to refine risk stratification and inform inclusive, evidence-based guidelines for low-speed motor vehicle collisions.

Authors’ contributions

H.Y.L. was responsible for project administration, performed formal analysis, and contributed to writing the original draft and revising the manuscript. C.Y.K. contributed to project administration and data curation. K.H.L. conceptualized the study and acquired funding. O.H.K. contributed to the development of the methodology and performed validation. H.J.K. and B.Y.L. performed formal analysis. Additionally, H.J.K. contributed to the methodology by designing the MRI acquisition sequences and performed formal analysis through image interpretation. G.H.K. contributed to the methodology by designing collision test scenarios. N.H.K. conducted the collision experiments and contributed to data curation. E.K.B. performed formal analysis by conducting markerless 3D motion analysis. J.Y.K. contributed to visualization by extracting 2D images from 3D motion data and also participated in formal analysis. All authors read and approved the final manuscript.

Funding

This research was supported by the Korean Insurance Development Institute and also supported by the Technology Innovation Program (RS-2024-00507638, Development of an Airbag System to Protect Rear-Facing Passengers in High-Speed Collisions for Autonomous Vehicles) funded by the Ministry of Trade Industry & Energy (MOTIE, Korea).

Data availability

The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

All procedures were approved by the Research Ethics Committee of Yonsei University Wonju Severance Christian Hospital (IRB approval number: CR323046). This clinical study complied with the International Conference on Harmonization (ICH) Guidelines and the principles of the Declaration of Helsinki, and was conducted in accordance with the Korean Good Clinical Practice (KGCP) and related regulations, taking into account the rights and safety of the subjects. All participants provided written informed consent.

Consent for publication

All participants provided written informed consent for publication of their anonymized data.

Competing interests

The authors declare no competing interests.